When a `LookupImplWitness` instruction is created in a generic function
for a witness obtained from a `BindSymbolicName`, it stores the
`BindSymbolicName` as the query self type along with the interface it
obtained from it.
Later, when an argument is substituted into the `LookupImplWitness` in
deduction, and it is re-evaluated, the `BindSymbolicName` in the query
interface's specific arguments was being substituted, but the same
`BindSymbolicName` in the query self type was not. This was because we
did not substitute into the type of `BindSymbolicName`. In this case the
type is a `FacetType` which has inside it one or more specific
interfaces. The same substitution needs to be applied to both the
interfaces in the self type as to the interfaces in the query.
This issue was found by a fuzzer - though in a weirder and more invalid
way, by putting all of the code for our test case inside an `interface`,
which creates an implicit generic `Self` in the enclosing context and
yet makes it concrete inside a function body.
And update tests to clarify that we should be able to do lookup into a
runtime facet value for an associated constant if the FacetType itself
provides that constant (with a `where` clause), but should not be able
to if it does not.
This fixes a fuzzer-found crash.
A fuzzer found that converting a pointer to a `FunctionTypeWithSelfType`
to `type` does an impl lookup and ends up in TypeIterator with the
`FunctionTypeWithSelfType` being iterated over in the self type, where
we crash. It's a valid type, so the iterator should be able to walk over
and return it.
We also constructed a similar example for `FunctionType`.
* Add // === headlines on group of file shards.
* Rename shard files to shorten given the file context them and add
`todo` where appropriate.
Not splitting this file for now.
Part of #5150.
* Split to 4 files: `function_param_int16`, `function_param_int32`,
`function_param_unsupported.carbon`, `function_return`.
* Add // === headlines on group of file shards.
* Rename shard files to shorten them and make them more consistent given
the file context.
* Deduplicate identical .h files and group their tests.
Potential future improvements:
* Split further. For example, pointers and references might be somewhat
separate from int primitives.
* Remove `import_` shard file prefix, as it repeats itself, but leaving
for now as it makes it more explicit.
Part of #5263
I think this would probably have prevented the missed include in #5469
-- it would've just failed completely with a "missing prelude"
diagnostic.
Also note this excludes the included IR from output, because it's
probably low-value to print.
We were importing all impls as non-final, since we forgot to set the new
field when constructing the imported Impl. Adds a test that fails before
this PR, since the imported Impl is treated as non-final.
If there will be no in-scope instructions printed, have
`FormatScopeIfUsed` skip the relevant scope.
Note, this only affects constants and imports. It's not changing the
file scope, which is usually printed when empty.
Right now, a lot of tests have started setting `--no-dump-sem-ir`. My
thought is that we can look at:
1. Put ranges in a bunch more files.
2. Shift more towards `--dump-sem-ir-ranges=only` instead of
`--no-dump-sem-ir`, because it allows mixing fail-tests with no IR
alongside tests that contain IR.
3. Evaluate switching the default to `--dump-sem-ir-ranges=only`, and
instead set `--dump-sem-ir-ranges=if-present` only in files that want to
typically show all IR (particularly import-related tests, where ranges
don't work well).
In real-world use, my thought is also that it'd be helpful to be able to
add the dump range comments to files, see the output (i.e., the default
behavior of `if-present`) but then also be able to pass `ignore` in
order to see the full IR without modifying the file (possibly also
useful in tests). That model is why I went for tri-state handling.
Note `only` can also have an interesting side-effect. Because core files
(including min_prelude versions) typically won't have ranges, they'd be
implicitly excluded.
Such impls will never be used, so they should not exist. And test that a
final impl partially overlapping a non-final impl is accepted.
There is a question about a final impl partially overlapping a final
impl that is part of
https://github.com/carbon-language/carbon-lang/pull/5337
Non-type values were being represented as "Concrete" in the type
structure's shape, which was incorrect when the value was symbolic. Now
they are represented as "Concrete" or "Symbolic" as appropriate.
When concrete, a matching concrete ConstantId is stored in the type
structure's concrete values, so that they will be used for equality
comparison. And then impl matching comparison is taught to look in the
concrete values for mismatches when it finds a "Concrete" or
"ConcreteOpenParen" shape on both sides of the comparison, and return
false if they are not equal. This reduces the number of candidate impls
selected in impl lookup, which avoids doing type deduction against impls
that won't match anyway due to the concrete values in the query not
matching concrete values in the impl. As a result we see fewer specifics
being generated in the semir.
The TypeStructureBuilder recursively iterates through a type, interface,
or facet value and constructs a type structure that includes each
concrete and symbolic value found. This iteration is a more general
thing that can be useful elsewhere. For instance, to get just the root
type out of a general type, it is the result of the first iteration
step.
We abstract out the iteration logic into a SemIR::TypeIterator to create
a clear boundary between the work of iterating and the work of building
the TypeStructure from it.
Along the way this pointed out some issues in the TypeStructureBuilder
where it could have ambiguity between types that include non-type
values. So we add some tests for these cases, and they now pass. There
are also TODOs left behind, as concrete TypeStructure is overly specific
right now in order to keep these tests passing, which means that the
concrete elements can't be used for impl lookup matching yet. Only the
shape of concrete vs symbolic is used for now, and then type deduction
is used to compare the actual concrete types, which could be skipped
when the concrete values could be compared directly and reject an impl
for not matching.
Instead of building the definition of a thunk immediately when we
generate the thunk declaration, wait until we reach the `}` of the
outermost class, interface, etc. -- at the same time when we would parse
the definition of the thunk if it were defined inline.
This fixes issues where we fail to define the thunk because it requires
an enclosing class to be complete, or its definition depends on
something declared later in the enclosing class.
Make the representation of a suspended function scope, and its
constituent suspended components, be move-only, and switch to passing it
around by rvalue reference instead of by value because it's expensive
both to move and especially to copy.
---------
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
This is fixing a deduce crash, with regression tests added in
binding_pattern.carbon. In `needs_substitution`, it adds the
`BindSymbolicName` with the compile time bind index corresponding to the
wrong generic scope, which causes a bad result. It appears
`needs_substitution` logic is no longer needed (per zygoloid,
`CheckDeductionIsComplete` handles related issues) so can be removed.
This changes the order of IR in use_assoc_const.carbon but the result
appears equivalent to me.
This was a fuzzer-found crash.
The canonical location of the instruction may be an entirely different
instruction, which the instruction in question was not imported from. In
particular, we shouldn't assume that we can use the constant value of an
instruction that the *location* of an imported instruction refers to as
the constant value of the imported instruction.
The only time we should be looking at the `ImportIRInstId` for a `LocId`
is when determining its location in some other file.
Fixes a crash when importing thunks (which can contain instructions
whose location points to an instruction in a differnt IR).
---------
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
When two non-final impls have the same type structure (neither is a
specialization of the other), it is an error unless they are within a
`match_first` block. For now, we don't have `match_first` implemented,
so it's always an error.
---------
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
A final impl must be written in the same file as the root self type or
the interface. This provides tests that should fail but don't yet for
writing a final impl in a third file that defines neither, as well as a
blanket final impl over an interface outside the file that defines the
interface.
When we fill the witness table with errors, set the witness id to an
error too, which signals to impl lookups to not use the impl.
Make the use of the `Impl` from the store more consistent once it's been
added to the store (or known to be there already).
- Adds parent information as a single-pass calculation
- Moves off `GetSubtreeTokenRange` because it's O(N)
- Stops using the node's subtree when formatting a single instruction
- This excludes `%F.call` and the following `return`, for example,
because only parameters are marked for formatting
When checking whether we can use the function in an impl directly to
satisfy a signature in an interface, allow the parameter names to differ
between the two declarations.
Instead of building an eval block as a separate pass at the end of a
generic, build the eval block incrementally.
The larger change here is that asking for the type or constant value of
an instruction now always returns an unattached type or constant value,
in order to preserve the behavior that we previously achieved by doing
the rewrite to attached types and constant values at the end of handling
the generic.
This also incidentally fixes some subtle issues where attached types and
constant values would leak out into check and cause it to get confused
about differences between attached and unattached values. Check should
no longer see attached values except where it explicitly asks for them.
---------
Co-authored-by: Dana Jansens <danakj@orodu.net>
Don't import `ImplWitnessTable` into the `constants` block, because we
generally don't put `Unique` constants there. This matches the handling
of the other kinds of `Unique` constants. In order to keep the
instruction visible in formatted SemIR, add it to the `imports` block
instead.
Also fix a bug in the instruction formatter that resulted in
instructions in the `imports` block being omitted from the output if
they were only referenced by earlier instructions in the `imports` block
and by instructions in the `constants` block. This was already resulting
in some referenced instructions being omitted from the output, but also
occurred frequently for `impl_witness_table` instructions after this
change because it is common for the only reference to those instructions
to be from `impl_witness` instructions in the `constants` block.
This prints instructions that are inside the range, and entities that
overlap with the range. Note this can lead to incomplete printing of
entity contents.
Once a concrete result has been found, it's not legal to write an `impl`
that would change the concrete result afterward.
---------
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
The location of the diagnostic is an instruction id, not a parse node.
Converting to a parse node to call ToTokenOnly will drop the descendents
of the instruction being diagnosed, but the diagnostic is about the
whole instruction not just the root parse node of whatever instruction
it happens to be.
This was originally needed to support constant evaluation of name
expressions, but that's now done in a different way.
This is actually a step toward treating all patterns as constants. The
upcoming change will do so in a slightly different way, and so it will
simplify the review to start from a baseline where patterns are never
constant.
Not entirely sure what the SemIR representation for this should be - do
we put the bogus thing in the vtable, and just not lower it later? The
patch currently doesn't add the function to the SemIR vtable - which
then means you could find a virtual function that's not in the vtable,
which seems similarly confusing.
I guess we could make the function non-virtual?
---------
Co-authored-by: Jon Ross-Perkins <jperkins@google.com>
Adds an empty `min_prelude/destroy.carbon` in anticipation of turning it
into an interface. Update `no_prelude` tests to be `min_prelude` and
import it where needed; in some cases, modify the file to remove the
dependency (i.e., rewrite code to have nothing to destruct).
Right now, return_scope_stack is being used to determine whether logic
is in a function scope. However, we need to handle nested entities
inside function scopes. For example where this crashes right now:
```
base class C(B:! bool) {}
fn F() {
class B {
extend base: C(true or false);
}
}
```
This is doing a few things to make this kind of code not crash:
- Split `scope_stack().Push` into `PushForDeclName`, `PushForEntity`,
`PushForExpr`, and `PushForFunction` so that better decisions can be
made about behaviors.
- Hide `return_scope_stack` in the API, instead using interfaces to get
at the underlying data.
- Also using `PushForFunction` to update it similar to the other stacks
that `ScopeStack` manages.
- Add `IsInFunctionScope` as the best way to determine presence in
function scope.
- Remove `PeekIsLexicalScope` since destruction really wants function
scope information anyways.
- Clean up `destroy_id_stack` handling to be for function scopes rather
than lexical scopes.
- Return after related `context.TODO`s in a couple more spots, so that
code doesn't proceed to add control flow in spite of the lack of
support.
---------
Co-authored-by: Geoff Romer <gromer@google.com>
We eliminate the `FacetAccessWitness` instruction, which would sometimes
immediately evaluate to a concrete `ImplWitness`, and sometimes remain
symbolic. This instruction is now replaced by `LookupImplWitness` in all
cases. To support the same use cases, when it is evaluated,
`LookupImplWitness` will look in the self value if it's a facet value,
and attempt to return a concrete `ImplWitness` from it before looking
for an `impl` statement.
The `LookupImplWitness` instruction's value is now canonical, even when
it evaluates to a symbolic `LookupImplWitness` instruction, by
canonicalizing the self value of the lookup query. This canonicalization
unwraps `FacetAccessType` and `FacetValue` instructions to get to an
underlying canonical facet value. However we must preserve and use the
non-canonical query while evaluating the instruction in order to look
for a concrete `ImplWitness` if the query self value was a concrete
`FacetValue`. The canonicalization ensures that symbolic witnesses
obtained from a facet value are compatible with those obtained from an
impl statement, as long as the self types originate from the same
canonical facet value though they may have been narrowed.
Member access now unconditionally does a `LookupImplWitness()`
operation, instead of only sometimes doing the lookup for a final impl
declaration.
`EvalImplLookupResult` is marked `[[nodiscard]]` so that we don't
construct it and forget to return it. This was a mistake made at one
point during the creation of this PR. And the `has_concrete_value()`
method no longer has a precondition that `has_value()` is true, since we
want to look for a concrete result only in the new use of
`EvalImplLookupResult` returned from lookup into the query self facet
value.
The TODO from `FacetAccessWitness` evaluation is addressed by ensuring
the index of the witness in the `FacetValue` comes from the required
interfaces of the `FacetValue`'s type, and that the type (a `FacetType`)
is the same facet type used in the query to construct the `FacetValue`'s
witness block. This is made possible by eliminating the
`FacetAccessWitness` indirection. The lookup into a `FacetValue` happens
while evaluating `LookupImplWitness` and it does so directly on the self
value. This gives a consistent view of the witness set and the facet
type, as they both come from the same instruction.
All of this with 400 less lines of code. :)
---------
Co-authored-by: josh11b <15258583+josh11b@users.noreply.github.com>
Co-authored-by: Richard Smith <richard@metafoo.co.uk>
In a class, an `impl as` can now be both `final` and `extend` instead of
only one or the other.
In https://github.com/carbon-language/carbon-lang/issues/5319 we decided
this is already allowed by the design but was an oversight in the
implementation.
A `MetaInstId` is intended to represent a handle to an instruction in a
generic as an operand to a template action; substituting into the action
should not substitute into the referenced instruction.
Fixing this exposed a bug in `GetOrAddInst` where it would return an
`InstId` of an unattached symbolic constant in some cases, rather than
the `InstId` of an instruction that has the relevant (attached) constant
value. That's fixed for now by turning off the `GetOrAddInst`
optimization in that case, but in future we can refine this by adding
the instruction to the eval block for the generic only, and not to the
body of the generic.
Provide builtins for compound assignments instead of defining them in
the prelude as a use of a binary operator and an assignment. This allows
us to lower compound assignment directly to LLVM operations instead of
producing a function call. In the short term this also allows us to
define a type-generic compound assignment in the prelude.
Instead of evaluating a non-parameterized class or interface to a
constant with `SpecificId::None`, use the self specific for that class
or interface, which will not be `None` if there is an enclosing generic.
We already had conversion in place to implicitly convert these literals
to `type`. Now they can also convert to `FacetType`. This is done by
first doing a conversion to `type` and then converting that type value
to `FacetType`.